BAAP 2004 Colloquium: Poster presentations

Yolanda Vazquez Alvarez, N.Hewlett & N. Zharkova (Queen Margaret Univ. College, Edin.): The “trough effect”: an ultrasound study

Some researchers have reported a lowering of the tongue during the bilabial consonant of, for example, /ipi/. This “trough effect” has been studied using X-ray, EMG, EPG and acoustic data. However, there are limitations with these measurement techniques. In this study we use ultrasound to test for the presence of a trough. A total of 10 British English speakers were recorded pronouncing several examples of symmetrical ViCVi sequences with the bilabial stops /b p/ five times each, both as nonsense syllables and as sequences within real speech. This investigation seeks firstly to confirm that tongue shape during the bilabial C varies systematically according to the identity of the V and secondly to determine whether a systematic deviation of the tongue shape occurs during the consonant between V1 and V2.

Patricia Ashby (University of Westminster): Phonetics in one word: learning through fieldwork

Following a recent LTSN workshop on resources and techniques for teaching linguistics, I was motivated to devise a new approach to the assessment of phonetic knowledge in a first year undergraduate module in general and English phonetics. This approach, which I am calling “Phonetics in One Word”, encourages learners to observe speech and take responsibility for demonstrating their growing knowledge of phonetic subject matter from day one of their course. They are, effectively, doing fieldwork: making preliminary descriptions which they revise, correct and develop on a weekly basis, form “hypotheses” and, eventually, draw conclusions. Each learner is given a two- or three-syllable English word (a different word for every student: brambles, candyfloss, plate-rack, etc.). Each word offers a certain amount of phonetic interest – interesting consonants and/or consonant sequences, at least two different vowels, plus resultant coarticulatory potential. Notebooks are written up weekly, applying knowledge gleaned from both lectures and practical classes and demonstrating a growing grasp of the subject. This has proved extremely successful, illuminating the learning process, focusing the teaching process and enabling immediate application of skills as well as meaningful assessment of both skills and knowledge of theory. Students report a sense of ownership and achievement.

Bill Barry (University of Saarbrücken): Changing English vowel quality – looking beyond the Queen

The intra-speaker comparison by Harrington et al. (2000) avoided many of the methodo-logical problems normally to be expected when comparing data over time. Inter-speaker comparison of instrumental data is itself a problem since anatomical differences are confounded with the differences found from one period of time to another; the relation between raw formant values and the perceived vowel quality in natural productions is still an unsolved problem.

This study examines the vowels of Chamberlain, MacMillan and Blair, taken from speeches to the Nation, the UN and to Parliament and the Nation and compares them in three ways:

  1. The raw formant data in Hz. and auditorily transformed.
  2. The normalised vowel systems, taking the indiviuals vowel centroid and the F1 and F2 dispersion into consideration (cf. Barry et al. 1989).
  3. The auditorily determined (relative) positions of corresponding vowel pho-neme realisations within the IPA-Cardinal Vowel system.

The results are discussed in terms of the relationship between perceived and measured vowel qualities, and with regard to the similarities and differences between these and Harrington’s findings as well as those given for modern Standard Southern British English (Deterding 1997),

Luke van Buuren (University of Amsterdam): Changing English vowel quality – looking beyond the Queen

My Observations on Phonation (JIPA 1983) offered a theory of phonation based on a collection of some 200 ‘funny voices’ gathered from records, radio and television. It also shows a few of the photographs of my larynx ‘doing’ some of these funny things taken through a rigid fibre-optic bundle after earlier experiments with a laryngoscopic mirror. Finally, it mentions (some of the earliest and therefore black-and-white) video-recordings of my larynx made in 1979 by our late colleague Tony Anthony through a flexible fibre-optic bundle inserted through the nose. All this material was further supplemented by good video-recordings in colour taken through my mouth at Groningen University Hospital in 1985.

Being in a position, at long last, to digitalize and edit all this material on a Macintosh computer I am now in the process of putting together a dvd with the provisional title (after Manuel Garcia (1855)) Observations on the Human Voice, indeed including that famous-but-difficult-to-find article itself as well as some subsequent highlights.

I am hoping to show (parts of) the trial dvd to colleagues interested in the subject as a ‘poster-presentation’ for comments and especially for suggestions on how to make good videos of one’s own larynx oneself, i.e. independently of medical facilities and personnel.

Traci Curl (University of York): The phonetic details of spontaneous speech: a parametric analysis of “I don’t know”

Over recent years there has been an increase in the amount of attention paid to the phonetic details of spontaneous speech data. A significant proportion of this work has been conducted over very large corpora, in many cases in order to drive a variety of speech technology devices (e.g. speech recognition and speech synthesis systems). In this paper I suggest that it is timely to rethink our strategies for approaching the phonetic details of spontaneous speech, principally by not being selective in the details which we engage in documenting. The paper presents a detailed parametric phonetic analysis of a number of instances of the utterance “I don’t know”, drawn from a single informal telephone call. The analysis shows that there is striking variability in participants’ deployment of phonetic resources in this utterance. The paper concludes by relating these findings to other, qualitative analyses of the phonetics of spontaneous speech, in which an effort has been made to explore the relationship between fine phonetic details and interactional functions of talk.

John H. Esling, Allison Benner & Lisa Bettany (University of Victoria): Phonetic articulatory control in prebabbling

We will present preliminary results of a project whose objectives are to document the phonetic development of infants in the first year of life, to highlight the role of laryngeal mechanisms in phonetic development and to determine whether laryngeal and pharyngeal articulations develop differently according to whether or not the infant’s ambient language includes these sounds. We will describe (1) the collaborative web-based XML database and analysis network that we have set up for researchers in Victoria, Paris, Morocco and China and (2) some initial phonetic results from a number of infants in the survey to illustrate the development of pharyngeal constriction, airstream mechanism control, stop control, phonation type and pitch control in prebabbling.

Snefrid Holm (University of Trondheim): Acoustic differences between read and spontaneous Norwegian speech: support for a theory of individual strategies

This is a study of acoustic differences between read and spontaneous speech in Norwegian. The investigated parameters are F0-mean, F0-dynamics, intensity-dynamics and formants. 42 speakers are used. All parameters except formants display differences between the read and the spontaneous speech, but effects are present only for some parameters. Moreover, some effects are small and statistically significant only for subgroups of the speakers. Because of these uncertain results, a small extra investigation is made. This control-experiment shows the same kind of uncertain results as the first experiment.

This could be interpreted as the investigated parameters not being important for the difference between read and spontaneous speech. Nonetheless, most of the speakers are consistent with their own strategy when changing between the styles. A speaker may be consistent in that he uses F0-mean consistently whereas another speaker may use F0-dynamics consistently. The speakers not only use different parameters, but can also use the same parameter in different ways such that one person may have a higher F0-mean reading whereas another may have a higher F0-mean when speaking spontaneously. Thus the speakers have individual strategies when changing between read and spontaneous speech.

Similar studies by other investigators on read and spontaneous speech show a great diversity in both results and conclusion. Studies on this topic often use very few (often 1-3) speakers. Evidence for the theory of individual strategies can only be observed when studying the behaviour of many speakers.

Mark Jones (University of Cambridge): “New” rhotics in English: an acoustic study of labiodental /r/

The class of rhotic sounds (comprising approximants, fricatives, and trills, all at various places of articulation) presents problems of definition. A low third formant (F3) has been proposed as an acoustic characterisation, but does not apply universally. A labiodental approximant realisation of /r/ is increasingly reported among younger speakers of British English. Little acoustic work has been done on labiodental approximants, and it is unclear how to characterise them as potentially rhotic sounds. An acoustic study comparing labiodental and lingual approximant /r/’s found that the labiodentals lack a low F3 (Docherty and Foulkes 2001). It also remains unclear how labiodental /r/ might be contrasted with sounds like /w/. This study presents controlled acoustic data on labiodental /r/ and /w/ from two speakers of Southern British English. The findings indicate that labiodental /r/ does indeed lack a low F3. However, it is also apparent that labiodental /r/ shows a movement of the second formant (F2) to a frequency around that of F2-F3 in non-labiodental realisations of /r/. The F2 movement suggests a lingual articulation. The implications of this finding for contrast between labiodental /r/ and /w/, the acoustic characterisation of rhotic sounds and transcriptional practices are discussed.

Jonathan Midgley & Sarah Hawkins (University of Cambridge): Formant frequencies of RP monophthongs in four age-groups of men

Change underway in the vowel system of RP is investigated by comparing the frequencies of the first two formants of eleven monophthongs spoken in ??V?? contexts by five male speakers in each of four different age-groups: 20-25, 35-40, 50-55, and 65-73. The eleven words, together with nineteen filler words chosen to distract attention from the purpose of the experiment, were randomised four times and read by each speaker in citation form, for a total of 11 vowels x 4 repetitions x 5 speakers = 220 items in each of the four age-groups (880 items total). F1 and F2 frequencies were measured in Hz and erb-rate. As expected, /?? and ??? are lowering, and ??? is fronting. Young speakers also front ?U?????????? /O/ and ??? seem stable. There is some evidence that the oldest age-group to show change in a vowel’s quality has particularly large individual variation, as if some individuals adopt a ‘conservative’ quality while others adopt a ‘progressive’ form. Such so-called “break groups” have implications for theoretical explanations of sound change. Possible future developments of RP monophthongs are discussed.

Katerina Nicholaidis, Jan Edwards, Mary Beckman & Georgios Tserdanelis (University of Thessaloniki, Ohio State University): Acquisition of lingual obstruents by Greek children

This paper investigates the acquisition of lingual obstruents in Greek. While Greek has a rich system of lingual obstruents, there has been limited research on their acquisition. Previous research on English and German has shown that coronal stops are acquired before dorsal stops and it has been suggested that this may be a universal pattern (Smit et al., 1990). However, recent research observed the opposite pattern for data from Japanese. Yoneyama, Beckman and Edwards (submitted) proposed that these differences are due to phoneme frequency variation across languages. In English, for example, /t/ is more frequent than /k/ and is acquired earlier, while in Japanese the opposite is the case. Yoneyama et al also pointed out that the difference between Japanese and English may be related to the fact that front vowels are more frequent than back vowels in English while the reverse is true in Japanese. Research by MacNeilage et al (2000) has suggested that there is a physiological preference for CV combinations in which the consonant and vowel have the same place of articulation. This work would predict earlier acquisition of coronal obstruents in a language with more front vowels and earlier acquisition of dorsal obstruents in a language with more back vowels. Of course, phoneme frequency is not the only factor that influences order of acquisition. In many languages, /s/ is a very high frequency sound, but is one of the latest acquired and most frequently misarticulated segments, e.g., Smit et al (1990). Frequency data on Greek has shown that it is like Japanese in that /k/ is more frequent than /t/. However, Greek is similar to English in that front vowels are more frequent than back vowels. Thus, Greek is an ideal language to investigate the influence of consonant and vowel frequency on consonant acquisition. In this study, three groups of children (two-year-olds, three-year-olds, four-year olds) produced the sounds /t/, /k/, /s/, /x/ in all five vowel contexts (/a, e, i, o, u/) in word-initial position in familiar words. Their productions were tape-recorded and transcribed. Order of acquisition and error patterns were examined and compared to data from other languages.

WITHDRAWN

Leendert Plug (University of York): Reduction patterns in context: the case of eigenlijk

This paper describes the phonetics of the word eigenlijk in a corpus of unscripted Dutch conversation. Previous work has suggested that the phonetics of eigenlijk, commonly glossed as ‘in fact’ or ‘actually’, are highly variable, and has attempted an account of the variability in terms of articulatory reduction. This paper shows that the variability is even more extensive than has been previously assumed, and is difficult to explain in terms of across-the-board temporal compression or gestural smoothing. Furthermore, a consideration of tokens of eigenlijk in context reveals a complex picture of interacting factors, including speaker identity, local speech rate, syntactic position, and sequential environment, influencing the phonetics. Implications for phonetic and phonological approaches to reduction patterns are discussed. It is argued that while most previous accounts of reduction patterns have been derivational, attributing special formal status to canonical forms, findings such as those of the present study are better accommodated in a declarative model of the phonetics-phonology interface.

Anna-Maija Rist (University of Aberdeen): VOT patterns in Finnish-English bilinguals’ productions of /p t k/

This paper reports on analysis of VOT data from 5 Finnish-English bilinguals resident in Hertfordshire, England. VOT measurements were made of word-initial /p t k/ in both languages. The speech production patterns of the bilingual speakers are contrasted with monolingual Finnish and English VOT values reported by Suomi (1980) and Lisker & Abramson (1964) respectively. Data from comparably-aged monolingual controls are also presented. Preliminary results indicate that while the bilingual speakers have clearly acquired separate production patterns for VOT in Finnish and English, their VOT values in both languages are slightly higher than those used by monolingual speakers in either language.

This pattern suggests that while the bilinguals maintain VOT-based distinctions between Finnish /p t k/ and English /p t k/, the contrasts are implemented differently from those used by Finnish and English monolinguals. The phonological implications of these findings will be discussed, making reference to Khattab’s (2002) findings for Arabic-English bilingual children.

Catherine Sangster (BBC Pronunciation Research Unit): The work of the BBC Pronunciation Research Unit

The three phoneticians who make up the BBC Pronunciation Research Unit are responsible for researching, transcribing and delivering pronunciations in all languages to the BBC’s broadcasters. This poster will explain how this work is carried out, discuss the phonetic processes involved, and consider some of the complicating factors.

I will discuss the challenges posed by the need to provide accessible, easy-to-read pronunciation advice to people who cannot read IPA and who are not necessarily native speakers of English. I will illustrate the advantages and disadvantages of our current approach (“BBC Modified Spelling”) and offer a phonetician’s perspective.

In addition, I will consider changing approaches to anglicisation. Broadcasters increasingly wish to give a more international impression, and this is often reflected in the abandonment of established anglicised pronunciations of place names and personal names in favour of forms closer to the original language. This means either that non-English phonemes must be introduced into an established transcription system (which is based on English spelling conventions and which was developed with monolingual speakers of English in mind), or that existing parts of the transcription system must be redeployed to approximate the non-English sounds which are now required.

James Scobbie (Queen Margaret University College, Edinburgh): Prevoiced, short lag, long lag: recurrent but not universal categories

The idea that there is a universal set of phonetic/phonological features becomes harder to defend the more it is discovered just how much language-specific phonetic detail there is. A recent study of Korean concludes: “it would take a great deal of procrustean effort to force Korean stops into the categories that have been developed for phonological descriptions of other languages” (Cho, Jun, Ladefoged 2002:222). Such cross-linguistic comparisons are, however, confounded by systematic differences in the phonological system which are themselves very likely to influence fine phonetic detail.

In this study, I examine phonetic variation in VOT within a single language (English), indeed in a single isolated community (the Shetland Isles). The speakers have very similar (or identical) phonologies of the stop system, though their accents differ. The results demonstrate the flexibility which individual speakers with different parental backgrounds bring to the specification of their own phonetic targets for VOT. The speakers no more make a clear unambiguous use of the traditional three categories than Korean does. Yet their pooled behaviour does provide some evidence of prevoiced, short lag and long lag VOT targets. Rather than being universal, however, such categories may recur, somewhat flexibly, due to shared functional underpinnings of contrast.

Linda Shockey (University of Reading): Long/short, tense/lax, [±ATR]?

Many textbook descriptions of Standard Southern British and General American English take it for granted that English has two sets of vowels which can be appropriately described as ‘long’ and ‘short’. While authors largely agree on the membership of these sets for SSB, there is some disagreement amongst those describing GA. Some mention it only in passing and do not regard it as a significant aspect of the vowel system. This paper looks at the long/short distinction from the perspectives of history, phonology, and language teaching and invites discussion about whether these terms are useful or misleading.

Rachel Smith (University of Cambridge): The role of allophonic detail in word spotting

This poster examines the role of systematically-varying fine phonetic detail in the segmentation of continuous speech into words by native English listeners. It is well known that listeners use information about the phonology of their native language to identify word boundaries. Because systematic variation in fine phonetic detail reflects structural linguistic organization of all types-grammatical, morphological, lexical, and phonological-phonetic detail could play an important role in segmenting the speech signal. I describe a word-spotting experiment that investigated the effect of allophonic mismatch on word segmentation. The primary manipulated variable was whether or not nonsense sequences containing an embedded real word were spoken as containing a word boundary, or with no boundary (e.g. oath in PUZ # oath, PUZoath, puz # OATH and puzOATH, where # denotes a word boundary and capitals denote stress). Listeners spotted words faster and more accurately when sequences were spoken as two words rather than one. This effect was modulated by the stress pattern of the sequences, suggesting sensitivity to the phonetic reflexes of complex phonological structure. I explore some acoustic properties underlying the perceptual results are explored, and suggest possible mechanisms for processing fine phonetic detail in word segmentation and recognition.

Rosalind Temple (University of York): “-t,d deletion” in York English, or is it?

At BAAP in 2002, Temple & Tagliamonte presented the results of a variationist analysis of the well-known sociolinguistic variable known as “-t/d deletion”, that is the variable deletion of word-final coronal stops in CC# clusters. We showed that our York English data did not confirm the patterns which are widely acknowledged in studies of “-t/d deletion” to be relatively consistent across dialects of English world-wide.

This paper presents some preliminary findings of a study of the same data (that is, words with possible final CC[+cor]# clusters), starting from a phonetician’s perspective and setting aside the assumptions on which variationist accounts rest. The data are variable in that different tokens of the same word in the same context are pronounced with and without perceptible reflexes of final alveolar stops, but the assumption that a uniform variable deletion rule can account for patterns of behaviour across all word-final consonant clusters (with or without regular morphological constraints) may not be the most constructive starting point.

Dominic Watt and Carmen Llamas (University of Aberdeen): Variation in the Middlesbrough English vowel system

In this paper we present findings from an investigation of variation in the vowel system of Middlesbrough English. The properties of English in Middlesbrough – and Teesside generally – have received virtually no attention in the linguistic and phonetic literature, and to our knowledge the Middlesbrough vowel system has never been described in any formal, systematic way. We pay particular attention here to the acoustic characteristics of the FACE, GOAT and NURSE variables (Wells 1982), vowels which have been found to be particularly sensitive to speaker age, social class and gender within the speech of a sample of speakers from the nearby Tyneside conurbation (Watt 2000; Watt & Milroy 1999). We suggest that the patterns observed in the data are indicative of accent levelling towards a northern English regiolect (see Watt 2002; Britain 2002), and link them to Llamas’ (2000) findings for consonantal variables in Middlesbrough English.

Maria Wolters, Nicholas Miller & Ghada Khattab (Newcastle University & QMUC): An efficient annotation scheme for Alternate Motion Rate data

Alternate motion rate (AMR) tasks provide invaluable information about the speech motor abilities of a person. In these tasks, subjects are asked to repeat a CV sequence (where C is an (un)voiced plosive) as quickly as they can for several seconds (typically five). AMR tasks test not only interarticulator coordination and articulatory precision, but also the coordination between laryngeal and supra-laryngeal subsystems. Three types of measures have been proposed for AMR data:
articulation rate (syllables per second),
duration measures (consonant and vowel durations,
duration of voicing)
energy measures (energy during plosive closure)

In this talk, we present an annotation scheme and a set of Praat scripts for analysing AMR data which were developed for a large-scale project focussing on the speech of persons with hypokinetic dysarthria secondary to Parkinson’s Disease. Our annotation scheme combines all three types of measures and allows for fast analysis of large quantities of data. We will evaluate the scheme in terms of interrater reliability and present data from a comparative analysis of 50 Parkinson’s patients and 50 healthy controls.

Irena Yanushevskaya (Trinity College Dublin): Acoustic correlates of the syllable-cut in Hiberno-English

The paper is concerned with the acoustic correlates of the syllable cut in Hiberno-English. An attempt has been made to find correlation between the number and placement of intensity peaks within the vocalic nucleus of the syllable and the auditory impression created by vowels belonging to subsets of checked (short) and free (long) vowels. The results proved no such correlation, the major feature responsible for the auditory impression of a tighter VC coupling most likely being the duration of the vowel.